Rich descriptor framework to text generation using graphs and structural neural encoders
View Patent ↗Technology for using a bi-directed graph convolutional neural network (“BGCNN”) to convert RDF data into natural language text. Some embodiments perform RDF-to-Text generation by learning graph-augmented structural neural encoders, consisting of: (a) bidirected graph-based meta-paths encoder; (b) bidirected graph convolutional networks encoder, and (c) separated attention mechanism for combining encoders and decoder to translate RDF triplets to natural language description.
1. A computer-implemented method (CIM) comprising:
training a bi-directed graph convolutional neural network (BGCNN) using a plurality of training data sets;
receiving a resource description framework (RDF) data set including computer readable information corresponding to a plurality of RDF triples;
creating, by machine logic, a bidirected graph data set that includes a bidirected graph inclusive of all of the information of the plurality of RDF triples; and
translating, using the BGCNN, the bidirected graph into a piece of natural language text.
2. The CIM of claim 1 wherein the training includes at least one of the following: supervised learning and/or unsupervised learning.
3. The CIM of claim 1 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
encoding, by a meta-paths encoder module of the BGCNN, meta-paths associated with the bidirected graph.
4. The CIM of claim 3 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
encoding, by a convolutional network (CN) encoder portion of the BGCNN, the bidirected graph.
5. The CIM of claim 4 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
combining, by a separated attention mechanism portion of the BGCNN, at least one encoder and at least one decoder to translate the RDF data set to the piece of natural language text.
6. The CIM of claim 1 wherein the BGCNN includes:
a bi-graph-based-meta-paths encoder;
a graph convolutional network encoder input; and
a bi-graph-convolutional-network encoder.
7. A computer program product (CPP) comprising:
a set of at least one storage devices; and
computer code stored collectively in the set of storage devices, with the computer code including data and instructions to cause a processor set to perform at least the following operations:
training a bi-directed graph convolutional neural network (BGCNN) using a plurality of training data sets,
receiving a resource description framework (RDF) data set including computer readable information corresponding to a plurality of RDF triples,
creating, by machine logic, a bidirected graph data set that includes a bidirected graph inclusive of all of the information of the plurality of RDF triples, and
translating, using the BGCNN, the bidirected graph into a piece of natural language text.
8. The CPP of claim 7 wherein the training includes at least one of the following: supervised learning and/or unsupervised learning.
9. The CPP of claim 7 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
encoding, by a meta-paths encoder module of the BGCNN, meta-paths associated with the bidirected graph.
10. The CPP of claim 9 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
encoding, by a convolutional network (CN) encoder portion of the BGCNN, the bidirected graph.
11. The CPP of claim 10 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
combining, by a separated attention mechanism portion of the BGCNN, at least one encoder and at least one decoder to translate the RDF data set to the piece of natural language text.
12. The CPP of claim 7 wherein the BGCNN includes:
a bi-graph-based-meta-paths encoder;
a graph convolutional network encoder input; and
a bi-graph-convolutional-network encoder.
13. A computer system (CS) comprising:
a processor set;
a set of at least one storage device; and
computer code stored collectively in the set of at least one storage device, with the computer code including data and instructions to cause the processor set to perform at least the following operations:
training a bi-directed graph convolutional neural network (BGCNN) using a plurality of training data sets,
receiving a resource description framework (RDF) data set including computer readable information corresponding to a plurality of RDF triples,
creating, by machine logic, a bidirected graph data set that includes a bidirected graph inclusive of all of the information of the plurality of RDF triples, and
translating, using the BGCNN, the bidirected graph into a piece of natural language text.
14. The CS of claim 13 wherein the training includes at least one of the following: supervised learning and/or unsupervised learning.
15. The CS of claim 13 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
encoding, by a meta-paths encoder module of the BGCNN, meta-paths associated with the bidirected graph.
16. The CS of claim 15 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
encoding, by a convolutional network (CN) encoder portion of the BGCNN, the bidirected graph.
17. The CS of claim 16 wherein the translation of the bidirected graph into a piece of natural language text includes the following sub-operation:
combining, by a separated attention mechanism portion of the BGCNN, at least one encoder and at least one decoder to translate the RDF data set to the piece of natural language text.
18. The CS of claim 13 wherein the BGCNN includes:
a bi-graph-based-meta-paths encoder;
a graph convolutional network encoder input; and
a bi-graph-convolutional-network encoder.